Methylation analysis as a diagnostic tool for sweat gland tumours classification with emphasis on the distinction between digital papillary adenocarcinoma and mimickers

Abstract

Aims

Digital papillary adenocarcinoma (DPA) is a rare sweat gland carcinoma arising on acral sites. The main differential diagnosis included tubular adenoma, hidradenoma, poroid hidradenoma, and mixed tumours, distinction between DPA and these mimickers being crucial for therapeutic management. Recently, HPV42 was identified as the main oncogenic driver of most DPA. Controversially, a few sweat gland tumour cases diagnosed as “DPA” but lacking the HPV42 genome and harbouring instead a BRAF ^V600E mutation, a genetic hallmark of tubular adenomas, have been recently described. Methylation analysis is a powerful tool for tumour classification systems. In this context, the aim of the present study is to evaluate the accuracy of methylation analysis for sweat gland tumour classification with special emphasis on the distinction between DPA and mimickers.

Methods and results

Twelve DPA, 11 tubular adenomas, 12 hidradenomas, 8 apocrine and 6 eccrine mixed tumours, 7 poromas and 6 adnexal sweat gland carcinoma not otherwise specified (NOS) were submitted for DNA methylation profiling. The results of this analysis show that most of these tumour types formed their own unique cluster, setting them apart from the others. In particular, DPA cases clustered together and were distinct from other tumour entities including tubular adenomas.

Conclusions

Our data support the distinction between DPA and tubular adenomas as two unique entities and further confirm DNA methylation profiling as a relevant tool for tumour classification.

Abbreviations

DNA

deoxyribonucleic acid

DPA

digital papillary adenocarcinoma

HPV42

human papillomavirus 42

HRM

high resolution melting

NOS

not otherwise specified

Introduction

Adnexal tumours are epithelial neoplasms arising from the skin and harbouring follicular, sebaceous, or sweat gland differentiation. ¹ Close to 50 tumour entities are described in the 5th edition of the World Health Organization classification of skin tumours, most of them being sweat gland tumours. For this latter group, precise case classification may prove difficult, especially in small biopsies. In this setting, despite the current classification system being largely based on morphology, specific recurring genetic events characteristic of particular tumour entities have been identified in recent years and may be employed for diagnostic purposes as ancillary tools in difficult‐to‐classify cases. ¹

An illustrative example is digital papillary adenocarcinoma (DPA), a rare type of sweat gland carcinoma that predominantly affects acral skin. ² The primary differential diagnoses include other sweat gland tumours that occur on acral sites, such as tubular adenoma, hidradenoma, poroid hidradenoma, and mixed tumour. ³ Distinction between DPA and mimickers is crucial for therapeutic management, as metastatic spread occurs in approximately 30% of DPA cases, whereas other simulators typically exhibit indolent behaviour.

Recent studies have identified HPV42 as the primary oncogenic driver of DPA. ⁴ , ⁵ Two large studies, including one conducted by our group, demonstrated that detection of HPV42 is a sensitive and specific diagnostic tool in the diagnosis of DPA. HPV42 detection is mutually exclusive with other oncogenic drivers present in mimickers such as CRTC1/3::MAML2 fusion in hidradenoma, YAP1::NUTM1 fusion in poroid hidradenoma, and BRAF ^V600E mutation in tubular adenoma. Controversially, two recent papers suggested the presence of BRAF mutations in a subset of DPA lacking HPV42 genome and harbouring low‐grade morphological features. ⁶ , ⁷ , ⁸ , ⁹ However, it remains contentious whether these cases should be classified as DPA or tubular adenoma occurring on acral sites, and this issue has yet to be conclusively determined. ¹⁰ To complicate matters further, another published series reported a CRTC3::MAML2 fusion in one case of DPA and a TRPS1::PLAG1 fusion in a further case, both of which were associated with the HPV42 genome. ¹¹

DNA methylation profiling is expected to be preserved during tumour cell oncogenesis and might therefore be indicative of the nature of the cell of origin in cancers. ¹² Therefore, DNA methylation analysis is a powerful tool for tumour classification. ¹² In this context, the aim of the present study is to evaluate the accuracy of methylation analysis for sweat gland tumour classification with special emphasis on the distinction between DPA and other sweat gland tumours.

Methods

Case Selection

A total of 62 adnexal sweat gland tumours were identified from the consultation's files of the authors (TK, NM, PS, EC, MB) and among the CARADERM network. Only primary tumours were considered for inclusion. The histological and genetic characteristics of some of these cases (i.e. DPA, n = 6), tubular adenoma (n = 4), hidradenoma (n = 6) and mixed tumours (n = 14) have been reported in prior studies. ³ , ¹³ The design of this retrospective study was in agreement with the requirements for the use of biological material in research proposed by our institutional ethics guidelines (Local Ethics Committee in Human Research, Tours, France; no. ID RCB2009‐A01056‐51).

HPV42 Genome Detection

Total DNA was isolated from formalin‐fixed and paraffin‐embedded (FFPE) tissue samples using the Maxwell 16 instrument (Promega, Madison, USA) with the Maxwell 16 FFPE Plus LEV DNA purification kit (Promega). Quantitative detection of HPV42 genomic material was performed using quantitative PCR (qPCR) with primers and probe specific for the L1 open reading frame. The human gene albumin gene (ALB) was used as the normalizing reference.

BRAF Mutation Screening

HRM screening for BRAF mutations was performed on a Light Cycler 480 II (Roche). Data were analysed with the Light Cycler 480 SW1.5 software. Pyrosequencing was performed on a PyroMark Q24 (Qiagen) according to the manufacturer's instructions.

RNA Sequencing

Next‐generation sequencing libraries were prepared with anchored multiplex PCR‐based methodology using the Archer^® FusionPlex^® Sarcoma Panel (Archer DX. Boulder, CO, USA).

Analysis of DNA Methylation

Genomic DNA was extracted from formalin‐fixed paraffin‐embedded tissues using Maxwell apparatus (Promega, Madison, WI, USA) with Maxwell 16 FFPE Plus LEV DNA Purification kits.

Methylation profiling employed the MethylationEPIC BeadChip (Illumina, San Diego, SA, USA). 300 ng of DNA was processed according to the manufacturer's protocol. Data were generated in the Department of Neuropathology of the University Hospital Heidelberg. Computational analyses were based on R version 4.6.1 (https://www.R‐project.org). t‐SNE plot was generated using the 10.000 most variable CpG‐sites upon standard deviation, 3.000 iterations and a perplexity setting of 10.

Results

Twelve DPA, 11 tubular adenomas, 12 hidradenomas, 8 apocrine and 6 eccrine mixed tumours, 7 poromas and 6 adnexal sweat gland carcinoma NOS were included in the present study (Figure 1). Median age was 58 (ranges 26–98), 36 patients were male, tumours were located on the head and neck (n = 15), on the trunk (n = 11), on the upper limbs (n = 19), on the lower limbs (n = 14) (anatomic location unknown for 3 cases). Interestingly, among tubular adenomas, 6 arose on acral sites. One of these was large and ulcerated.

Figure 1.

Morphologic features of digital papillary adenocarcinoma and mimickers. A DPA, a tubular adenoma, a hidradenoma and an apocrine mixed tumour are shown. DPA harbours a solid and cystic architecture which in this specimen is associated with prominent papillary formation. This tumour is composed of luminal cells forming glands associated with a population of myoepithelial cells. Tubular adenoma is characterized by multiple dermal tubules arranged in individual units and separated by collagen. Granular eosinophilic material is present in the glandular lumina. Hidradenoma is composed of lobules of cells with clear or eosinophilic cytoplasm surrounded by hyalinized stroma. Although not a constant feature, presence of goblet cells is a useful diagnostic clue. Apcorine mixed tumour consists in a proliferation of eosinophilic cells freqnetly harbouring plasmocytoid appearance embedded in a fribrohyaline and myxoid stroma.

Genetic findings were obtained in 43 cases. Consistent with prior descriptions, the HPV42 genome was detected in all instances of DPA (12 out of 12). Additionally, CRTC1::MAML2 fusions were identified in hidradenomas (10 out of 12), BRAF V600E mutations in tubular adenomas (9 out of 11), and TRPS1::PLAG1 fusions in the apocrine variant of mixed tumours, respectively (5 out of 8).

DNA methylation analysis of the entire set of these tumours (Figure 2) revealed that most of these tumour types formed their own unique cluster, setting them apart from the others. It is especially of note that the eccrine variant of mixed tumour was separate from cases of apocrine mixed tumours, as suggested by earlier transcriptomic analysis. ¹³ In contrast, cases of sweat gland carcinoma NOS appeared to segregate into two distinct groups. Of particular importance, DPA cases also formed a distinct cluster separate from tubular adenomas, hidradenomas, and mixed tumours, with the exception of one DPA case found adjacent to a heterogeneous group of sweat gland tumours including hidradenoma, and tubular adenoma. This neoplasm arose on the second finger of the right hand in a 61‐yearyear‐old man and harboured all prototypic morphological features of DPA. No mutation or significant fusion transcript was detected in this case. No recurrence was observed after a 26‐month follow‐up.

Figure 2.

Methylation profiling of sweat gland tumours. t‐SNE analysis is depicted. Such analysis. Demonstrate that most tumour types display a unique cluster, setting them apart from the others. Notably, DPA cases also formed a distinct cluster distinct from tubular adenomas, hidradenomas, and mixed tumours, with the exception of one DPA case with low tumour cellularity.

Discussion

In this study, we have shown that most sweat gland tumour entities, including hidradenoma, poroma, eccrine and apocrine variants of mixed tumours, tubular adenoma and DPA, exhibit specific methylation profiles. Thus, although additional studies are required for validation, our results provide proof of the proposal that methylation profiling may be useful in classifying skin tumours. This innovative tool could be particularly valuable in identifying new tumour entities or refining diagnostic criteria for existing ones.

In this context, a controversial question is whether sweat gland tumours arising on acral sites, harbouring BRAF mutations and lacking HPV42 genomes should be classified as DPA (“low grade”) or tubular adenomas. Indeed, seven cases diagnosed as DPA and harbouring BRAF mutations have been recently described in the literature. ⁶ , ⁷ , ⁹ , ¹¹

While the authors assert that “it is now accepted that DPAs cannot be distinguished from benign adenomas based on histological features”, ⁷ our assessment based on the histopathological images provided suggests that these tumours exhibit prototypic morphological features characteristic of tubular adenomas, which are distinctly different from those typically associated with DPA. ³ , ¹⁰

Notably, tubular adenomas consist of individual tubular structures separated by collagen, while DPA frequent solid areas with back‐toto‐back glands (Figure 1). Of note, tubular adenomas may display large size, poor demarcation, luminal granular eosinophilic material that sometimes mimics necrosis, and increased mitotic activity. However, in tubular adenomas, none of these features correlate with aggressive behaviour, and while some cases have been reported as adenocarcinoma in situ, no tumours have behaved in an aggressive manner. ⁶ , ¹⁴ Accordingly, none of the BRAF‐mutated DPA cases described in the literature have demonstrated aggressive behaviour. This discrepancy highlights the potential value of molecular characterization in conjunction with morphology as an aid in the resolution of this issue. In this context, methylation profiling further suggested that tubular adenomas and DPA represent two separate tumour entities with distinct morphologic, genetic, epigenetic, and prognostic features. This distinction is crucial for patient management, as surgery with wide margins and potentially sentinel lymph node biopsy ¹⁵ would be the treatment of choice for DPA cases but not for tubular adenomas. In line with this view, our data also demonstrated that DPA, hidradenomas, and apocrine mixed tumours have distinct methylation profiles. In this context, it is interesting that on the illustrations provided for the two cases diagnosed as “DPA” but harbouring CRTC3::MAML2 or TRPS1::PLAG1 fusion reported in the literature, ¹¹ both tumours had typical morphologic features of hidradenoma and mixed tumour respectively (see figure 1B ¹¹ ) with the presence of goblet cells in the case with CRTC3::MAML2 fusion and plasmacytoid eosinophilic cells in the other.

In our cohort, one DPA case with no specific clinical or morphological features clustered near a heterogeneous group of sweat gland tumours including hidradenomas and tubular adenomas; these findings could be attributed to the low cellularity of this sample (less than 20% of tumour cells) or a limitation of the procedure and highlight that DNA methylation profiling needs to be interpreted carefully, together with clinical and histopathological findings.

In conclusion, our data support the distinction between DPA and tubular adenomas as two separate entities, challenge the notion of BRAF ^V600E mutated DPA and further confirm that DNA methylation profiling is a useful tool in tumour classification.

Author Contributions

PS, NM, AVD, and TK performed study concept and design, performed development of methodology and writing, review and revision of the paper; PS, NM, AT, CM, MB, EC, AVD, and TK provided acquisition, analysis, and interpretation of data, and statistical analysis; all authors read and approved the final paper.

Funding information

Illumina Grant.

Conflict of interest

All authors have no conflict of interest to disclose.

Institutional review board

The local Ethics Committee in Human Research of Tours (France) approved the study (no. ID RCB2009‐A01056‐51).

Data availability statement

Dataset of the present manuscript would be available on request.